preparatory distributed cortical synchronization ...gbrewer1/http___~gbrewer1... · preparatory...

Preparatory distributed cortical synchronization determinesexecution of some, but not all, future intentions

JUSTIN B. KNIGHT,a RICHARD L. MARSH,a,* GENE A. BREWER,b and BRETT A. CLEMENTZa

aBioImaging Research Center & Department of Psychology, University of Georgia, Athens, Georgia, USAbDepartment of Psychology, Arizona State University, Tempe, Arizona, USA*deceased

Abstract

Associating intentions to events that cue future behaviors is a central aspect of human cognition. There is limitedunderstanding of the neural dynamics supporting recognition of intention-related events, with little known about howpre-event brain state varies as a function of intention specificity. Prior to recognized events (that cued planned behavior)occurring during an unrelated activity, we found increased steady-state visual evoked potential (ssVEP) and intrinsicgamma synchronization for ill- compared to well-specified events, as measured by EEG. Enhanced fronto-temporo-parietal ssVEP synchrony emerged preceding ill compared to well-specified events, and the degree of synchronypredicted the completion of ill-specified intentions but predicted failure to complete well-specified intentions. Distinctexecutive processing and neural states are therefore optimal for anticipating and fulfilling future intentions varying inevent specificity.

Descriptors: Cognitive control, Gamma, Oscillatory synchrony, Prospective memory, Steady-state visual evokedpotential

The coordination of attentional and mnemonic processes thought tosubserve future-oriented, intention-related cognition is termed pro-spective memory (PM). Planned intentions (e.g., buy a birthdaycard) can be associated with well-specified events (e.g., the campusbookstore) or ill-specified events (e.g., any location that sells cards)which, when noticed, instigate execution of the intention. Impair-ments of PM in aging (Henry and MacLeod, 2004) and clinicalconditions (Foster, McDaniel, Repovš, & Hershey, 2009; Woods,Twamley, Dawson, Narvaez, & Jeste, 2007) are reduced or nonex-istent when well-specified events are associated to intentions. Awell-specified event can be a specific word (e.g., horse) that par-ticipants are explicitly aware they should respond to with theintended action (e.g., a special keypress) when the word occurs inan unrelated task. However, an ill-specified event is typically morecategorical in nature and participants are to respond with theintended action when they encounter, for instance, an animal word.Thus, the event is not explicitly specified as it could be any wordfrom the animal category (Hicks, Marsh, & Cook, 2005; Marsh,Hicks, Cook, Hansen, & Pallos, 2003; Meeks & Marsh, 2010).Behavioral studies suggest that different attentional and mnemonicprocesses support noticing well- versus ill-specified events (Hickset al., 2005), but whether neural mechanisms underlying PM

co-vary with event type is unknown. The current paper examinesthe neural dynamics present directly prior to successful prospectiveremembering to elucidate how the brain prepares for detecting bothwell- and ill-specified events associated with future intentions.

Cognitive processes supporting intention-related event recogni-tion are associated with activity in the rostrolateral prefrontal cortex(rlPFC), lateral parietal and temporal cortices, and medial occipitalcortex (Burgess, Quayle, & Frith, 2001; Gilbert, 2011; Reynolds,West, & Braver, 2009). Gilbert (2011) demonstrated increased func-tional coupling of the rlPFC with posterior regions when partici-pants were to maintain the intention and detect intention-relatedevents embedded in an unrelated task. Accordingly, successful PMmay rely on frontal control processes (Miller & Cohen, 2001) thatmodulate other brain regions responsible for maintaining the inten-tion and/or for processing incoming information to identify eventoccurrences (Knight, Ethridge, Marsh, & Clementz, 2010; Reynoldset al., 2009; Simons, Schölvinck, Gilbert, Frith, & Burgess, 2006).Whether these top-down modulations are required regardless of thespecificity of the event type is unknown.

Two prominent theories yield competing predictions aboutwhether top-down attentional processes are needed to detect bothtypes of events. The preparatory attentional and memory processes(PAM) theory proposes that people must always rely upon atten-tional processes to detect events and fulfill intentions associated tothem (Smith, Hunt, McVay, & McConnell, 2007). By contrast, themultiprocess view (MPV) predicts that events can in some cases benoticed automatically (Scullin, McDaniel, & Einstein, 2010), forexample, when the event is salient, well-specified, and/or a strongassociation can be formed between the specific event and intended

This research was supported by the John and Mary Franklin Foundationthrough a fellowship awarded to J.B.K. We thank Benjamin Peck, AnastasiaBobilev, and Matthew Epperson for their assistance with data collection.

Address correspondence to: Brett A. Clementz, Department of Psychol-ogy, University of Georgia, 125 Baldwin Street, Athens, GA 30602-3013.E-mail: [email protected]

bs_b

s_ba

nner

Psychophysiology, 49 (2012), 1155–1167. Wiley Periodicals, Inc. Printed in the USA.Copyright © 2012 Society for Psychophysiological ResearchDOI: 10.1111/j.1469-8986.2012.01400.x

1155

action (Einstein et al., 2005). If none of these stipulations are met,which is the case for ill-specified intentions, then MPV asserts thatpreparatory attentional processes will be needed to notice events.For the well-specified intention in the present study, the specificword serving as the PM event was repeated across the unrelatedongoing task; this repetition could also influence the attentionalprocesses engaged to notice events. Repetition of the well-specifiedevent could strengthen the association of the event and intendedaction as the task progressed, perhaps stimulating reliance on moreautomatic processes to detect events, in line with predictionsof MPV but not PAM theory. The extent to which similar or dis-tinct neural dynamics are associated with fulfilling well- versusill-specified intentions, therefore, has important theoreticalimplications.

When preparatory attentional control processes are relied upon,they are thought to be engaged across the task to support a readi-ness to process incoming stimuli as possible intention-relatedevents. Extant neuroimaging investigations have associated thesePM attentional processes with rlPFC and parietal activations sus-tained across blocks of trials (Reynolds et al., 2009; Simons et al.,2006) and frontal event-related potential (ERP) modulations ontrials in which an event did not occur (Chen, Huang, Yang, Ren, &Yue, 2007; West, Bowry, & Krompinger, 2006). Behavioral workindicates preparatory attention is most prominent in the periodprior to PM events that are detected and elicit the intended action(West, Krompinger, & Bowry, 2005). Brain activity directly pre-ceding PM events that are noticed remains uncharacterized andwarrants investigation considering behavioral PM work and thevital influence of prestimulus brain states on attention (Weissman,Roberts, Visscher, & Woldorff, 2006), encoding (Otten, Quayle,Akram, Ditewig, & Rugg, 2006), cognitive flexibility (Leber, Turk-Browne, & Chun, 2008), and response anticipation (Hamm,Dyckman, Ethridge, McDowell, & Clementz, 2010).

Using dense-array electroencephalography (EEG), we exam-ined prestimulus steady-state visual evoked potentials (ssVEPs),which are oscillatory responses elicited by flickering stimuli(Regan, 1989). Steady-state VEPs (a) have a high signal-to-noiseratio, (b) provide a continuous neural response measure to pre-sented stimuli, (c) can be facilitated or suppressed via PFC medi-ated top-down control mechanisms, and (d) show enhancedsynchronization to the stimulus volley with attention (Andersen &Müller, 2010; Clementz et al., 2010; Kim, Grabowecky, Paller,Muthu, & Suzuki, 2007; Müller, Teder-Sälejärvi, & Hillyard,1998). Given these attributes and the well-characterized sensitivityof ssVEPs to cognitive control processes (for review, see Vialatte,Maurice, Dauwels, & Cichocki, 2010), the oscillatory dynamics ofssVEPs may provide a useful index of attentional control processesengaged prior to intention-related events. Considering the debatedrole of top-down control in PM, this approach offers valuableinformation that could further our understanding of the processesengaged in anticipation of an opportunity to fulfill an intention.Additionally, long-range synchrony of oscillations between frontalto posterior scalp regions increases with greater reliance on cogni-tive control (Sauseng et al., 2005). If preparatory attention isrequired to realize intentions associated with ill- but not well-specified events, we might expect enhanced ssVEP synchronizationto the relevant stimulus and/or enhanced long-range ssVEP syn-chrony preceding ill-specified events.

Additionally, synchronization of neural population activity inthe gamma frequency band (>30 Hz) increases in sensory cortical(Fries, Reynolds, Rorie, & Desimone, 2001) and parietal regions(Van Der Werf, Jensen, Fries, & Medendorp, 2008) as a function of

top-down control, serving to bias and maintain processing of task-relevant information (Fries, 2009). Investigation of pre-eventssVEPs provides information about evoked (time-locked) activityand its relation to PM; however, intrinsic oscillations (not time-locked to stimuli) in the gamma band also vary with cognitivecontrol and memory processes (Buzsáki, 2006). Examiningwhether gamma synchronization varies prior to successfullydetected ill- versus well-specified events should provide a morecomprehensive understanding of the preparatory brain state sup-porting PM. Here, we demonstrate an enhancement of evoked(ssVEP) and intrinsic (gamma) synchronization across a distrib-uted cortical network prior to recognized ill-specified, but not well-specified, events that led to successful prospective remembering.Furthermore, stronger frontal-to-temporal synchronization forssVEP oscillations directly prior to and during event recognitionpredicted PM performance (better for ill-specified and worse forwell-specified events).

Materials and Methods

Participants

Thirty right-handed participants (ages 18–22; 17 females) partici-pated to meet a research requirement. Participants providedinformed consent, displayed no signs of neurological impairment,were free of psychiatric and substance abuse disorders (self-report), and had normal or corrected-to-normal vision. This studywas approved by the University of Georgia Institutional ReviewBoard.

Stimuli

Lexical stimuli were presented using Presentation software on a21″ high-resolution monitor (60 Hz refresh) with participantsseated 70 cm away (see Figure 1). Each trial began with a smallwhite fixation square (5 cd/m2; .16″) appearing for 250 ms againsta dark background (0.5 cd./m2). A square-wave luminance modu-lated (100% depth) white rectangle (5 cd/m2; 2″ ¥ .5″) flickering at15 Hz was then presented over the square for 3,000 ms (prestimu-lus period). The square was removed, the flickering rectangleremained, and a luminance-modulated, linguistic stimulus flicker-ing at 15 Hz was superimposed and flickered in phase with therectangle for 1,500 ms (poststimulus period). The screen was thenblank for 1,500 ms to allow for settling of the ssVEP. The stimuliwere words, nonwords, or event-cues. Words and event-cues withmedium-to-high frequency of occurrence, 3–9 letters, and 1–3 syl-lables were acquired from the Kucera and Francis (1967) compen-dium. Words from the same compendium and with the samecharacteristics had 1–3 letters changed to create pronounceablenonwords (Knight et al., 2011).

Procedure

Participants were randomly assigned to either a well- or ill-specified event condition (n = 15 in each). The task was a 400-triallexical decision task (LDT), including 160 words, 200 nonwords,and 40 event-cues (henceforth referred to as events). On each trial,participants made a judgment about whether the randomly chosenstring of letters presented was a word or nonword. On a four-keyresponse box, participants pressed the “1” key with their left indexfinger if a nonword appeared and the “4” key with their right indexfinger if a word appeared. In line with previous studies and as is

1156 J.B. Knight et al.

common in PM research (Hicks et al., 2005; Knight et al., 2011,2010; Smith, 2003; Smith & Bayen, 2004), event trials comprised10% of the total trials. The type of event associated to the intentionwas manipulated between conditions. Participants in the well-specified condition received an intention associated to a specificword (i.e., the word horse, which occurred 40 times in the LDT).Those in the well-specified condition were told that if they encoun-tered the word horse during the LDT, they should press the “3” keyinstead of making a word response. In contrast, participants in theill-specified condition received an intention associated with asemantic category (i.e., animal words). Those in the ill-specifiedcondition were told to press the “3” key instead of making a wordresponse if they encountered an animal word (40 different animalwords were presented) during the LDT. Participants in both con-ditions were to complete the LDT and execute the same PMresponse identically; thus, only the nature of the event-cues dif-fered between conditions (ill-specified versus well-specified). Thisdesign follows from previous behavioral work, uses a similar pres-entation frequency of events relative to ongoing task trials (10%here), and, to anticipate, replicates the behavioral results from thosestudies (Hicks et al., 2005; Marsh et al., 2003; Meeks & Marsh,2010). After receiving PM instructions, participants completed abrief distractor task before beginning the LDT. A short breakoccurred half way through the task.

EEG Recording

EEG data were recorded vertex-referenced using a 256-sensorGeodesic Sensor Net and NetAmps 200 amplifiers (Electrical Geo-desics; EGI). The sensor net was adjusted until all pedestals wereproperly seated on the scalp, individual sensor impedances werebelow 50 kW, and there was no evidence of sensor bridging prior to

recording. Data were sampled at 500 Hz with an analog filterband-pass of 0.1–200 Hz using a Macintosh G4 running EGI’sNetstation software.

EEG Data Processing

Sensors around the neck and cheeks were excluded, leaving 211sensors. Raw data were visually inspected offline for bad sensorrecordings that were interpolated (<5% of sensors per subject)using a spherical spline method implemented in BESA 5.3 (MEGISSoftware). Data were transformed to an average reference anddigitally filtered from 1–50 Hz (12 db/octave rolloff, zero-phase).Artifact correction was achieved using the Independent ComponentAnalysis (ICA) toolbox in EEGLAB 9.0 (Delorme & Makeig,2004) running under Matlab (version 7.10, MathWorks). Independ-ent components with topographies representing saccades, blinks,heart rate artifact, and muscle artifact were removed. This ICAartifact correction technique has been shown to be able to reliablydecompose muscle artifact from brain activity in similar frequencyranges (Onton & Makeig, 2009). After artifact correction, an addi-tional down-sampled representation of the data was computed byinterpolating the sensor data to 27 sensors (International 10–20system) using spherical spline interpolation. This data reductiontechnique served to reduce the number of comparisons in initialexamination of scalp time-frequency data.

ssVEP Analysis

EEG responses to flickering stimuli included a Fourier componentthat matched the flickering frequency (i.e., first harmonic), as wellas harmonic components at two and three times the flickerfrequency (i.e., second and third harmonics, respectively; see

Figure 1. Trial sequence and spectral plots. A: Each trial was separated by a 1,500-ms intertrial interval (ITI). In both conditions, a trial began with a centralfixation square, which appeared alone for 250 ms. A white rectangle was then superimposed and flickered (square-wave luminance modulated) at 15 Hz for3,000 ms while participants attended to the center of the rectangle. The fixation square was then removed and a word, nonword, or event (well-specified eventdepicted) was superimposed and flickered with the rectangle for 1,500 ms, during which time participants responded in accordance with the lexical decisiontask. The event was either the word horse (well-specified condition) or an animal word (ill-specified condition). B: Spectral plots averaged over a set ofoccipital-parietal sensors (Oz, Pz, O1, O2, P3, P4, P7, and P8) depict that the EEG evoked harmonic components from 500–3,500 ms postflicker elicited bythe flickering rectangle in each condition were above noise. Grand average power (mV2; log-transformed for display purposes) is plotted.

Preparatory synchronization and prospective memory 1157

Figure 1). The presence of these harmonic components was statis-tically verified using the T2-circ statistic (Victor & Mast, 1991). Foreach participant in both conditions, the real and imaginary compo-nents of the Fourier transformed data were separately calculatedacross trials and subjected together to the circular t test. The result-ing T2-circ value when multiplied by n is F-distributed with 2n - 2degrees of freedom, where n equals number of trials included andreflects the consistency of the amplitude and phase of the oscilla-tory activity for a given frequency across trials with respect to thestimulus. This measure tests the extent to which evoked activity ata given frequency is present across trials, and it was conducted toensure ssVEP activity could be dissociated from noise before beingcompared between conditions. This analysis was completed at eachof the 27 sensors for frequencies 1–50 Hz (in .33-Hz steps) for anepoch spanning 500–3,500 ms postflicker (because ssVEPs typi-cally take ~ 400–500 ms to stabilize; Moratti, Clementz, Gao,Ortiz, & Keil, 2007) to determine the evoked activity that waspresent throughout the relevant flicker duration. Evoked activitywas considered significantly present if at least a third of the sensors(9/27) for more than half of the subjects in a given group (8/15)displayed T2-circ values at p < .05. Evoked activity at a givensensor was compared between conditions only if both conditionsmet these criteria. Evoked responses were observed for the higher-order harmonics, but they did not differ between conditions in anyanalyses; thus, these data are not presented.

Time-Frequency Quantification and Analysis

Oscillatory activities as a function of the type of event associated tothe intention were analyzed (i.e., well- versus ill-specified). Time-frequency transformations of the EEG data were computed from-500 preflicker to 4,000 ms postflicker (to ensure edge artifacts didnot distort the epoch of interest) by means of a modified Morletwavelet transform of single-trial data (1–50 Hz) using EEGLABtoolbox (Delorme & Makeig, 2004). The wavelet length increasedlinearly from .4–18 cycles for frequencies 1–50 Hz, offering theoptimal balance between temporal and frequency resolution at lowand high frequencies, respectively (Busch & VanRullen, 2010). Fora given trial k, the wavelet transform at each time point t andfrequency f produces a complex number in which the amplitudeand phase of the signal are represented by A and j, respectively:

A ek t fi k t f

( , )( , )ϕ

Spectral power was then computed for each time and frequencypoint from -250 ms preflicker to 3,500 ms postflicker (to avoidedge artifacts) separately for each trial by squaring the magnitudeof the resulting complex value. Power values were then averagedover trials for each participant, producing single trial power (Clem-entz et al., 2010; Moratti et al., 2007). In addition to examiningevent effects on single trial power for the ssVEP, we quantifiedactivity in the gamma band, excluding the evoked harmonics(gamma = 31–50 Hz).1

To characterize phase consistency of the ssVEP, the waveletcomplex output was normalized by amplitude for all time-

frequency points for each trial. The magnitude of this value aver-aged across trials produces the Rayleigh statistic, which is boundbetween 0 and 1 (with 1 indicating perfect phase-locking to thestimulus flicker) and will be referred to as intertrial coherence(ITC; Moratti et al., 2007). Due to the sensitivity of ITC to thenumber of trials contributed (i.e., ITC is positively biased withfewer trials or spectral estimates; Hipp, Engel, & Siegel, 2011), wecalculated ITC expected by chance given the number trials (Morattiet al., 2007):

Rchance = − 1

nclog( )

where Rchance is chance ITC, n is number of trials, and c is .5. Wesubtracted each participant’s chance ITC from the observed ITC.Resulting ITC values were Fisher Z transformed. ITC values forssVEP and power values for the ssVEP and gamma band weresubsequently averaged over epochs including the 250-ms preflickerepoch and 7 subsequent 500-ms epochs ranging from 0–3,500 mspostflicker for each sensor, participant, and condition. Initial analy-ses were conducted on these epoch bins of averaged activity.

Initial comparisons between conditions were conducted by con-structing a bootstrap estimate of the difference between the ill- andwell-specified conditions in power and ITC at each sensor (27-channel montage) and epoch for the ssVEP and gamma. A boot-strap sample for each condition was calculated by sampling (withreplacement) 15 participants’ oscillatory activity, resulting in twobootstrap samples (n = 15 for each). The mean difference betweenthese two samples was calculated for each sensor, epoch bin, andssVEP/gamma band. This process was repeated 10,000 times toobtain a bootstrap distribution of the differences between condi-tions. If the upper and lower 95th distribution percentile did notinclude 0, then the given difference was considered significant.These effects provided initial characterization of the temporal andscalp location difference and guided further analysis.

If either power or ITC value displayed significant effects atthree contiguous sensors during at least one epoch bin, then clusteraverages were formed using sensors from the dense-array montagethat corresponded to the scalp regions where initial analysesrevealed differences. These criteria were chosen to ensure analyseswere focused on effects that were sufficiently prominent across thescalp and time, in order to avoid potentially spurious effects that aredriven by a single sensor or time point. Activities in these sensorclusters were averaged over 50-ms bins spanning the entire trialperiod in order to more precisely characterize the temporal evolu-tion of the effects. These data split into 50-ms bins were analyzedusing a nonparametric cluster-based permutation test (Maris &Oostenveld, 2007) as implemented in FieldTrip toolbox (Oosten-veld, Fries, Maris, & Schoffelen, 2011). Traditional Neyman-Pearson approaches (e.g., Bonferroni correction) to the multiplecomparisons problem are inappropriately conservative with multi-variate brain activity data (Maris, 2012; Worsley, 2003).Permutation-based approaches, like the one used here, have beenadvocated as the best approach for controlling Type 1 error rates,when existing research does not offer data-driven predictions toreduce the data as is the case here (Maris, 2012). This test con-servatively controls for multiple comparisons by clustering neigh-boring time epochs that, when compared between conditions witha t test, display t values of similar magnitude and direction. Oncethe temporal clusters were determined, the sum of the t values wasused as the cluster-level statistic, and the cluster with the maximumstatistic was used as the test statistic that was compared to a

1. Given findings suggesting involvement of prestimulus theta(4–7 Hz) and alpha (8–12 Hz) activity with retrospective memory encoding(Fell et al., 2011) and cognitive processing efficiency (Klimesch, Sauseng,& Hanslmayr, 2007), power in these frequency bands was also examined.No differences in these frequency ranges emerged between conditions, sothese data are not presented.


randomization null distribution. To create this distribution, the datafrom all participants were randomly assigned to one of two groupsand the maximum cluster-level statistic was calculated 1,000 dif-ferent times. The proportion of randomized cluster-level statisticsthat exceeded the observed cluster-level statistic was used as the pvalue; significance level was set at p < .05. Due to the spatiallyextended effects observed initially for gamma power, data from all211 sensors with the same 50-ms epoch bins were analyzed withthe permutation test. The only difference here is that the clustersformed are based on temporal and spatial adjacency, with sensorswithin 3.2 cm considered as neighbors (~7 sensors).

Intersensor Coherence

The degree of ssVEP (15 Hz) synchronization between the 378sensor pairs (from the 27 sensor locations) was assessed and com-pared across conditions using intersensor phase coherence (ISC).ISC for each time point and sensor pair was computed by normal-izing the cross-spectrum from the two signals by their productamplitude. When averaged across trials, the magnitude of thecomplex output produces a value bound between 0 and 1, where 1indicates that two sensor signals maintain an identical difference inphase across trials. Using the same mathematical correctiondescribed above, ISC values were corrected for chance based on thenumber of trials for each participant. Then mean baseline ISC(-250–0 preflicker) was subtracted from each trial period timepoint because our primary interest was ISC that was linked to thestimulus flicker. Those values were Fisher Z transformed and aver-aged for 7 subsequent 500-ms epochs ranging from 0–3,500 mspostflicker for each sensor pair, participant, and condition. Above-baseline ISC (each condition mean > 0) at each sensor pair andepoch bin was analyzed between conditions with the bootstrapdifference test and using the upper and lower 99th percentile todetermine significance. Each sensor pair in each epoch demonstrat-ing a significant ISC difference was submitted to an intersubjectcorrelation analysis in which participants’ ISC strength was corre-lated with their proportion of successfully recognized events foreach condition.

ssVEP Source Localization

A modified linearly constrained minimum variance vector beam-former (Gross et al., 2001), as implemented in Brain ElectricalSource Analysis software (BESA 5.3; Megis Software), was usedto localize the primary neural generators of the 15 Hz activity. Thebeamformer can be used to localize a specific range of activity inthe time-frequency domain, and it has been implemented in manyprevious studies and shown to produce reliable solutions consistentwith other methods (Praamstra, Kourtis, Kwok, & Oostenveld,2006; Staudigl, Hanslmayr, & Bäuml, 2010). The beamformer usesthe cross-spectral density matrix (time-frequency equivalent to adata covariance matrix) calculated on single trials to estimate thechange in activity in the interval of interest relative to the baseline(or preflicker) period (-500–0 ms to equate time points in baselineand postflicker epochs). A realistic head model was constructedbased on a four-shell ellipsoid (Berg & Scherg, 1994), and the fourhomogeneous shells were warped into an ellipsoid that best fit the3D electrode coordinates for each participant. Standard conductivi-ties were estimated for the brain, cerebrospinal fluid (CSF), skull,and scalp (.33, 1, .0042, .33, respectively). A spatial filter wasapplied at each voxel (4 mm3) throughout the brain to estimate15 Hz (�.5 Hz) activity. Beamformer source analysis was per-

formed for each of the 7 postflicker 500-ms epochs (0–3,500 ms)for each participant and condition. Due to similarity in the sourceresults, beamformer solutions were combined for 0–1,000, 1,000–2,000, and 2,000–3,000 ms. The percentage of the maximum activ-ity was calculated for activity in each voxel in order to obtain thelocation of the generators that were primarily contributing to thescalp-recorded activity.

Results

Behavioral Performance

Well-specified events (M = .94, SE = .01) were correctly respondedto significantly more often than ill-specified events (M = .84,SE = .02), t(28) = 4.35, p < .001, d = 1.57, replicating previousbehavioral research (Hicks et al., 2005; Marsh et al., 2003).Latency to respond to cues that were noticed did not significantlydiffer between groups (well: M = 814 ms, SE = 23; ill: M = 888 ms,SE = 36), t(28) = -1.71, p = .1; though one may wonder whetherthere was enough power to detect this effect, this result replicatesprevious behavioral work in which a more powerful test of thiseffect was possible (i.e., n = 36 in each condition; Marsh et al.,2003). Word and nonword latencies and accuracy were analyzed inseparate mixed analyses of variance (ANOVAs) with factors lexi-cality (word, nonword) and condition (well, ill). Analysis of laten-cies only revealed a significant main effect of lexicality,F(1,28) = 8.8, p < .05, ηp

2 24= . , indicating the common finding thatwords (well: M = 733, SE = 145 vs. ill: M = 809, SE = 109) wereresponded to faster than nonwords (well: M = 778, SE = 165 vs. ill:M = 829, SE = 122). Accuracy analysis also only revealed a signifi-cant main effect of lexicality, F(1,28) = 5.6, p < .05, ηp

2 17= . ,whereby accuracy was lower for words (well: M = .9, SE = .02 vs.ill: M = .93, SE = .02) than nonwords (well: M = .91, SE = .03 vs.ill: M = .97, SE = .01), likely reflecting a speed-accuracy tradeoff.For word and nonword trials, no effects or interactions involvingcondition were significant. Overall, well-specified events werenoticed more often than ill-specified events; however, when theevents were noticed, the intended keypress was executed on asimilar time scale.

ssVEPs

The synchronization of neural responses underlying ssVEPspreceding noticed well- and ill-specified events (during the pres-timulus period) was examined using ITC, which provides anamplitude-independent measure of the degree to which neuralactivity is phase-locked to the stimulus flicker (Moratti et al.,2007). ITC at the driving frequency was significantly greater pre-ceding ill- compared to well-specified events at sensors over bilat-eral frontal regions (centered on F8 and F7; Figure 2). Thisincreased ITC was primarily observed during the 500-ms epochdirectly preceding the presentation of the cue. Steady-state VEPsingle trial power was investigated also, but no significant differ-ences emerged. Finding differences in ITC, but not single trialpower, is consistent with recent reports demonstrating that themajority of variance of ssVEPs in this frequency range is capturedby ITC, suggesting ssVEPs are largely driven by phase alignmentrather than amplitude augmentation (Moratti et al., 2007).

Follow-up analysis revealed that ITC was significantly greaterin advance of noticed ill-specified events than well-specified eventsat both frontal clusters from 2,600–2,900 ms, p < .05, and 2,550–2,750 ms, p < .05, for the left and right clusters respectively


(Figure 2). No baseline (preflicker) ITC differences were observed,revealing these effects were only present during the stimulusflicker. Thus, directly prior to successfully noticed events, therewas increased neural synchronization over the frontal cortex for ill-as compared to well-specified intentions.

Perhaps this effect did not result from differences betweenconditions in preparatory attention devoted to noticing events;rather, the occurrence of event repetitions in the well- but not theill-specified condition may have led to habituation processes andaccounted for the differences. This seems unlikely because theprimary epoch of interest occurred prestimulus when participantswere unaware that the upcoming stimulus would be an intention-related event, so stimulus-specific repetition responses could nothave influenced this activity. Nevertheless, perhaps the repeatedpresentation of the well-specified event across the task induced atype of global habituated processing of stimuli as the task pro-gressed. If such list-wide habituation processes occurred in thewell-specified condition and led to the observed differences, thenbrain activities elicited by words and nonwords should decreasefrom the beginning to the end of the task in the well-specifiedcondition (Schacter, Wig, & Stevens, 2007). ITC was computed foreach third of the task, and analyses revealed no such negative trendin activity across the task for either condition (see online Supple-mental Material). The observed differences, therefore, are moreparsimoniously accounted for by differential engagement of pre-paratory processes between conditions.

For poststimulus processing, baseline-corrected ITC analysisrevealed a desynchronization for the ill- compared to the well-specified condition that was primarily lateralized to the right frontalregion and consistent over the poststimulus interval (Figure 2), as

indicated by a significant Hemisphere ¥ Condition interaction,F(1,28) = 5.27, p < .05, ηp

2 16= . The scalp topographies (Figure 3)show that in the ill-specified condition ssVEP ITC gradually spreadfrom posterior to anterior scalp regions throughout the prestimulusperiod and reached maximal spread directly prior to event presen-tation. Such dispersion of ssVEP activity across the scalp suggestsanterior cortical regions were recruited, in addition to visual areas,and entrained to the stimulus flicker.

To characterize whether additional cortical regions were con-tributing to the scalp-recorded ssVEP activity in the ill- comparedwith the well-specified condition, we localized the primary neuralgenerators of 15 Hz activity (i.e., ssVEP driving frequency) using abeamformer analysis (see Methods). In the well-specified condi-tion, maximum 15 Hz activity was localized to medial occipitalcortex (15,-95,15; ~BA 18; average Talairach coordinates ofmaximal source across prestimulus interval), which remained rela-tively constant throughout the prestimulus period (Figure 3). Thislocalization is consistent with previous studies demonstrating aprimary visual cortex source for the ssVEP (Clementz, Wang, &Keil, 2008; Fawcett, Barnes, Hillebrand, & Singh, 2004). Simi-larly, the maximal 15 Hz activity during the initial 1,000 ms offlicker in the ill-specified condition was localized to the medialoccipital cortex (14,-92,25; ~BA 18/19). During latter intervals(1,000–3,000 ms) leading up to the event occurrence, however, thecortical sources of the maximal 15 Hz activity were more distrib-uted with activation in anterior left frontal regions (-48,-10,55;~BA 6/4; average coordinates across latter prestimulus intervals;Figure 3). During the poststimulus interval, the maximum 15 Hzactivation was localized to the occipital-parietal region in bothconditions (well: 14,-100,1; ill: 2,-96,17; ~BA 18).

Figure 2. Frontal ssVEP intertrial coherence. A: Time-frequency representations of intertrial coherence (ITC) for both conditions averaged over bilateralfrontal clusters highlighted in (B). The first and second dotted lines in all plots (A,C) represent flicker and event onset, respectively. B: Topography of thedifference between the ill- and well-specified conditions in ITC during the overlap of temporal epochs that were significantly different between conditionsfor the left and right frontal clusters (2,600–2,750 ms). Marked sensors reflect the left and right frontal clusters chosen based on initial analysis and overwhich ITC was averaged (separately) and compared between conditions throughout the trial period. C: Time-frequency representations of the differencebetween conditions for both sensor clusters. The black box highlights the pre-event temporal epoch during which ssVEP ITC was significantly enhanced forthe ill- relative to the well-specified condition for each cluster. Warmer colors indicate ill > well; cooler colors indicate well > ill. D: Baseline correctedpoststimulus ITC for left and right frontal clusters. Error bars reflect SEM.


Intrinsic Oscillatory Synchronization

We also investigated whether intrinsic oscillatory dynamics in thegamma frequency range (31–50 Hz) differed in preparation fornoticing and responding to well- versus ill-specified events. Usingthe same analysis approach described above, we compared gammasingle trial power between conditions after excluding activity atharmonic responses evoked by the stimulus flicker. Significantlyincreased single trial power in the gamma band (31–50 Hz) wasfound for the ill- compared to the well-specified condition overoccipital, parietal, temporal, and frontal regions throughout theentire trial period (Figure 4).

Given the temporally and spatially widespread extent of thegamma differences found in initial analyses, all sensor data span-ning the entire period were submitted to a cluster-based permuta-tion test to determine sensor time clusters that significantly differedbetween conditions (see Methods). This analysis revealed thatgamma activity was significantly increased in the ill- compared tothe well-specified condition over left parietal cortex from 1,250 ms

until 3,500 ms, p < .05 (Figure 4), indicating enhanced, parietallymediated, cognitive control (Fries, 2009) in preparation for notic-ing ill-specified events. Examination of poststimulus gamma activ-ity revealed a larger poststimulus gamma band desynchronizationin the ill-specified condition, with this effect reaching significance400–500 ms poststimulus, p < .05.

15 Hz Synchronization Between Sensors

Finally, we investigated synchronization of ssVEP activity acrossthe scalp by quantifying ISC, which measures the consistency ofphase relationships between pairs of sensors across trials. ISCbetween frontal and posterior sensors has been associated with themagnitude of cognitive control processes (Sauseng et al., 2005).Using a nonparametric bootstrap difference test, comparisons ofbaseline (preflicker) adjusted ISC averaged over successive 500-msepochs revealed more extensive ISC in the ill- compared to thewell-specified condition (all ps < .01; Figure 5). No sensor pairsdisplayed increased ISC for the well-specified condition. In con-

Figure 3. Topographic plots of ssVEP ITC and beamformer source localizations of 15 Hz activity (ssVEP driving frequency) across the trial period. All plotsdepict averaged activity over 3 subsequent 1,000-ms prestimulus epochs and the 500-ms poststimulus epoch. A,C: ssVEP ITC topographies for the well- andill-specified conditions, respectively. Notice different scales for pre- and poststimulus topographies. B,D: Beamformer source localizations of the 15 Hzactivity for the well- and ill-specified conditions, respectively. For each condition and epoch, source activity (% Max) plotted on a template brain (in Talairachspace). Axial slices are in radiological convention.


trast, numerous sensor pairs showed significant coherence in theill-specified condition, and these relationships progressivelyincreased across the prestimulus period with the most extensiveISC differences occurring directly prior to the event and during thepoststimulus interval. During the first 1,500 ms of stimulation,there was high temporal-frontal ISC, which was followed by pri-marily left parietal-temporal-frontal ISC from 1,500–2,500 ms andextensive bilateral frontal-parietal-temporal ISC throughout theremainder of the trial period (Figure 5).

These findings demonstrate that evoked synchronization acrossa distributed frontal-temporal-parietal network emerges to a greaterdegree in advance of recognized ill- than well-specified events.Furthermore, across participants in the ill-specified condition,frontal-temporal ISC directly preceding recognized ill-specifiedevents was positively correlated with the percentage of successfullyrecognized events, F8-A1: r = .56, p < .05. Poststimulus interhemi-spheric frontal ISC was also positively correlated with the percent-age of recognized ill-specified events, F8-F7: r = .52; F10-F7:r = .56, ps < .05 (Figure 6). In contrast, the well-specified condi-tion exhibited an inverse relationship, such that poststimulus inter-hemispheric frontal and frontotemporal ISC were negativelycorrelated with the percentage of recognized well-specified events,F8-F7: r = -.74; F10-F7: r = -.69; F10-T8: r = -.57; F4-A2:r = -.54, all ps < .05 (Figure 6). ISC in this network did not exhibitsuch relationships with performance measures for the word and

nonword trials, providing evidence that this synchrony is specifi-cally related to processes associated with remembering to completeprospective memories. These temporally ordered neurobehavioralrelationships demonstrate that greater neural population synchro-nization in this distributed network was predictive of better pro-spective remembering for ill-specified intentions but worseprospective remembering for well-specified intentions, revealing adouble dissociation of the brain states necessary to complete inten-tions varying in specificity.

Discussion

The present results provide unique characterizations of the brainstate directly prior to successful PM and indicate this preparatoryactivity varies as a function of the type of event that is associated tothe intention. In anticipation of ill- compared to well-specifiedevent occurrences, there was enhanced ssVEP synchronization andgamma band synchronization that was present over frontal andparietal regions. Moreover, ssVEP activity over frontal, temporal,and parietal regions exhibited enhanced inter-regional synchroni-zation in advance of recognized ill-specified events, and the degreeof this distributed synchronization predicted improved PM recog-nition for ill-specified events but impaired recognition for well-specified events.

Figure 4. Intrinsic oscillatory synchronization in the gamma (g) band pre- and poststimulus. A: Topography of average t values during the significanttemporal epoch (1,250–3,500 ms; p = .05) resulting from the comparison of gamma band activity between conditions. Marked sensors reflect the parietalcluster of sensors that were found significant during more than 50% of the time points during the significant epoch. B: Time-frequency difference mapresulting from comparing between conditions the mean single trial power for the marked sensors in (A). The black box highlights the epoch during whichgamma band activity significantly differed between conditions. The first and second dotted lines in the plot represent flicker and event onset, respectively.C: Baseline corrected (relative to 500-ms epoch preceding the event) poststimulus gamma band raw power (mV2; *p < .05). Data represented asmean � SEM. D: Depicts the temporal evolution of the t value topography and sensor cluster (marked sensors) for the significant epoch. Time labels referto the center of five 50-ms epochs that are evenly dispersed across the significant interval and are representative of the progression of the effect (3,275 msis poststimulus). Warmer colors indicate ill > well; cooler colors indicate well > ill.


Fig

ure

5.In

ters

enso

rco

here

nce

diff

eren

ces

betw

een

cond

ition

spl

otte

dfo

r7

subs

eque

nt50

0-m

sep

och

aver

ages

from

0–3,

500

ms

post

flick

er.S

enso

rna

mes

onax

esre

fer

toIn

tern

atio

nal1

0–20

syst

emlo

catio

ns.

For

each

epoc

h,th

edi

ffer

ence

betw

een

cond

ition

sin

base

line

corr

ecte

din

ters

enso

rco

here

nce

(ISC

)is

depi

cted

for

sens

orpa

irs

that

sign

ifica

ntly

diff

ered

betw

een

cond

ition

s(p

<.0

1).W

arm

erco

lors

indi

cate

ill>

wel

l;co

oler

colo

rsin

dica

tew

ell>

ill.S

enso

rpa

irs

with

ISC

that

disp

laye

dsi

gnifi

cant

corr

elat

ions

acro

sspa

rtic

ipan

tsw

ithth

epe

rcen

tage

ofev

ents

notic

edar

ehi

ghlig

hted

byci

rcle

s(p

<.0

5;se

eFi

gure

6).

Red

indi

cate

sth

atpa

irw

aspo

sitiv

ely

rela

ted

tope

rfor

man

cein

the

ill-s

peci

fied

cond

ition

,mag

enta

indi

cate

sth

atpa

irw

aspo

sitiv

ely

rela

ted

tope

rfor

man

cein

the

ill-s

peci

fied

cond

ition

and

nega

tivel

yre

late

dto

perf

orm

ance

inth

ew

ell-

spec

ified

cond

ition

,and

blue

indi

cate

sth

atpa

irw

asne

gativ

ely

rela

ted

tope

rfor

man

cein

the

wel

l-sp

ecifi

edco

nditi

on.


Evoked Synchronization in Prospective Memory

Steady-state VEP synchronization, as measured by ITC, spreadfrom posterior to anterior regions across the epoch in the ill-specified condition and was enhanced over the frontal scalp for theill- compared to the well-specified condition just prior to eventoccurrence. Data from the present and previous studies indicate theinitial neural source of the ssVEP is the occipital cortex (Clementzet al., 2008; Fawcett et al., 2004). Attentional modulations ofssVEPs in early visual areas are thought to enhance signal strengthof attended spatial locations or features by a sensory gain and/or aneural response synchronization mechanism (Andersen & Müller,2010; Kim et al., 2007). Throughout the prestimulus period, par-ticipants in both the well- and ill-specified conditions were attend-ing to the stimulus flicker location, explaining the absence ofssVEP differences over posterior scalp regions and the similaroccipital ssVEP generator found for both conditions during initialflicker epochs. During the latter part of the prestimulus flickerinterval in the ill-specified condition only, lateral frontal cortex

primarily contributed to ssVEP activity variance, which is consist-ent with reports of local and distributed ssVEP sources (Srinivasan,Bibi, & Numez, 2006). Further, the right frontal ssVEP desynchro-nization observed poststimulus for the ill- compared to the well-specified condition supports the notion that right frontal regionswere differentially entrained to the flicker, as entrained ssVEPgenerators are momentarily disrupted upon additions to the stimu-lus display (Moratti et al., 2007). This right hemisphere dominancefor the ill-specified condition may be similar to right hemispheredominance in attentional modulations (Shulman et al., 2010).These anterior cortical signals could result from thalamocorticalinput and/or corticocortical propagations from posterior regions(Srinivasan, Russell, Edelman, & Tononi, 1999). The lateral frontalregions recruited in advance of ill-specified events are perhapsassociated with top-down modulation of posterior cortical regions(e.g., Kastner & Ungerleider, 2000) responsible for processing theforthcoming stimulus given the particular context.

Involvement of the lateral frontal regions here is consistent withnumerous functional magnetic resonance imaging (fMRI) and posi-tron emission tomography (PET) studies that have observed greaterrostrolateral PFC activation during PM tasks as compared to taskswith no PM component (Burgess, Gonen-Yaacovi, & Volle, 2011;Burgess et al., 2001; Gilbert, 2011; Reynolds et al., 2009; Simonset al., 2006). One study reported that activation patterns in the PFCvaried according to the intended action one planned to execute (i.e.,whether one was intending to add or subtract subsequently presentednumbers; Haynes et al., 2007). Similarly, we found anticipatoryrecruitment of the frontal cortex that varied with the specificity of theintention-related event that one was preparing to recognize. Thisfinding provides evidence that activity in some frontal regions alsomay be sensitive to the nature of the event associated to the intendedaction (c.f. Gilbert, 2011). The notion that the enhanced frontalssVEP activity for ill-specified intentions reflects executive controlprocesses that subserve successful fulfillment of this type of inten-tion is consistent with a recent ERP study examining PM and taskswitching. West, Scolaro, and Bailey (2011) found an enhancedlateral frontal modulation in ill-specified PM blocks relative to taskblocks without a PM component, and this modulation was sensitiveto whether task switching was required for the upcoming trial,suggesting the frontal modulation was supporting control processesthat were interfered with by the need to reconfigure the task set.These previous studies have reported tonic frontal activity (e.g.,activation sustained across trials or slow wave potentials) associatedwith PM (Burgess et al., 2011; Reynolds et al., 2009; West, 2011;West et al., 2011); here, we provide evidence that such anticipatoryactivity is transiently enhanced prior to recognized events and thatthis activity varies with intention type. The fixed timing of theinterstimulus interval (ISI) may have had some influence on theobserved preparatory activity; however, investigations of prestimu-lus activity in retrospective memory that used fixed (Guderian,Schott, Richardson-Klavehn, & Düzel, 2009) or variable (Fell et al.,2011) ISIs reported similar effects of prestimulus oscillations onencoding that only varied slightly in their time of onset. Thus, thereported effects here are not easily explained as an artifact of the ISI.Future work should examine the influence of ISI on the observedpreparatory PM activity.

Role of Intrinsic Gamma Synchronization inProspective Memory

Gamma band activity (Fries, 2009) has been associated with per-ceptual mechanisms (Singer, 1999), visuospatial attention (Fries

Figure 6. Relationship of ISC to behavioral performance for bothconditions. Between-subject relationship of percentage of events noticed(i.e., that instigated the intended keypress) against prestimulus (A) andpoststimulus (B–D) ISC for sensor pairs with significant correlations ineach condition (all ps < .05). A: F8-A1 ISC directly prior to the recognizedevent (2,500–3,000 ms) was predictive of a higher percentage of eventsnoticed in the ill-specified condition (r = .56). B: Depicts the positiverelationship between poststimulus F8-F7 and F10-F7 ISC (3,000–3,500 ms) and performance for the ill-specified condition (r = .52; r = .56,respectively). C: Depicts the negative relationship between poststimulusF8-F7 and F10-F7 ISC (3,000–3,500 ms) and performance for thewell-specified condition (r = -74; r = -.69, respectively). D: For thewell-specified condition, the negative relationships of poststimulus F4-A2and F10-T8 ISC (3,000–3,500 ms) with behavioral performance are plotted(r = -.54; r = -57, respectively). In all plots, filled circles represent theill-specified condition and filled triangles represent the well-specifiedcondition.


et al., 2001), and working memory maintenance (Van Der Werfet al., 2008). In the present study, participants in both conditionsattended to the same flickering stimulus, yet those in the ill-specifiedcondition exhibited enhanced posterior gamma synchronization thatbegan ~1,700 ms prior to event presentation and remained through-out the trial period. This gamma enhancement, indicative of pre-cisely timed local neural spiking that maximizes the impact onpostsynaptic neuronal populations (Fries, 2009), could reflectincreased working memory maintenance of intention-related infor-mation in the ill-specified condition (e.g., animal category informa-tion and/or the intended keypress) and/or increased attentionalengagement (e.g., selecting for category-related information).

An attentional explanation predicts a continued increase ingamma synchronization once relevant animal words appeared.Gamma activity decreased, however, in the ill-specified conditionupon event presentation and was significantly lower than the well-specified condition after 400 ms poststimulus. This finding is con-sistent with recent reports of dissociations between pre- andpoststimulus activity (Fell et al., 2011) and suggests the processessupported by gamma oscillations may have been engaged pres-timulus in the ill-specified condition but poststimulus in the well-specified condition. Accordingly, the anticipatory parietal-occipitalgamma synchronization for the ill- compared to the well-specifiedcondition may reflect maintenance of conceptual representations ofthe intended keypress (Gilbert, 2011), which could be reflexivelyaccessed poststimulus via associative retrieval processes supportedby the hippocampus in the well-specified condition (Einstein et al.,2005; Moscovitch, 1994). Such reflexive or spontaneous retrieval,which is consistent with empirical and theoretical proposals (Ein-stein et al., 2005; Gordon, Shelton, Bugg, McDaniel, & Head,2011), could support the high levels of performance in the well-specified condition. Additionally, the parietal activity may reflectdifferential engagement of a retrieval mode that is a neurocognitiveset aimed at treating incoming stimuli as possible retrieval cues foran intended behavior (Guynn, 2003; Tulving, 2002). Parietal activ-ity (both fMRI localized and scalp ERPs) has been consistentlyreported in previous PM studies during periods of maintaining aPM intention while completing an unrelated task. Our resultsextend these findings by suggesting the locus of this parietal activ-ity associated to PM may be in low gamma synchronization.

Synchronization Across a Distributed Network

Intersensor phase coherence between distal EEG sensors reflectslong-range synchronization of neural population activity across thebrain, as opposed to the more local synchronization indexed by ITCat a given sensor cluster (Srinivasan et al., 1999). Long-range syn-chronization across distributed anterior and posterior brain regionswas enhanced for the ill- compared to the well-specified conditionprior to recognized events and became progressively more exten-sive leading up to event occurrence. Inter-regional synchronizationcould result from re-entrant modulations between corticocortical(or cortico-thalamo-cortical) projections and/or from simultaneousthalamic input to distinct cortical regions (Srinivasan et al., 1999).Synchronous thalamic input should produce synchronized ITCmeasurable at individual sensors over those regions (Srinivasanet al., 1999), which was not observed throughout the prestimulusinterval. Enhanced ISC in the ill-specified condition, therefore, ismore consistent with corticocortical (or cortico-thalamo-cortical)interactions of task-relevant brain regions. Though consistent withprevious fMRI findings implicating the involvement of PFC toposterior cortex connectivity in PM (Gilbert, 2011), the current

data further our understanding by (a) demonstrating the presenceand evolution of such a distributed network directly precedingevent recognition, (b) establishing that such a network is not uni-versally relied upon to fulfill varying types of intentions, and (c)providing initial evidence that synchronization across this networkbefore and during the event is predictive of behavioral indices ofPM performance.

Our results demonstrate a double dissociation of the depend-ence on this synchronized network as a function of the type ofintention. Not only was distributed synchronization enhanced inthe ill-specified condition, but this distributed synchrony and PMperformance were positively correlated for ill-specified intentionsand negatively correlated for well-specified intentions. Interre-gional oscillatory synchrony may provide precisely timed windowsof excitability that coordinate information transmission and com-munication between cortical regions (Hipp et al., 2011). Thepresent results reveal preparatory coordination of distributedregions that is beneficial for executing ill-specified intentions butmay interfere with completing well-specified intentions. Theinverse relationship found for well-specified intentions wouldbenefit from future experimental scrutiny due to the high level ofperformance in the well-specified condition. Nonetheless, thesedata contribute to the growing body of work implicating theinvolvement of interregional synchronization in various cognitiveprocesses (Hipp et al., 2011; Sauseng et al., 2005; Srinivasan et al.,1999) and possibly demonstrate dissociable roles of such synchro-nization in PM prior to and during event recognition.

An aspect of the current design worth considering as it relates toprevious work is the presentation frequency of the intention-relatedevents. Here, the event repeated across the task in the well- but notill-specified condition. We demonstrated that the strength of oscil-latory activity did not decrease across the task in the well-specifiedcondition, indicating it is unlikely that habituated processingoccurred (Schacter et al., 2007). Further allaying concerns thathabituation resulting from event repetition caused the currenteffects, studies examining repetition priming and habituationresponses have reported that such effects are associated withdecreased posterior gamma activity during the stimulus presenta-tion (Gruber, Malinowski, & Müller, 2004; Gruber & Müller, 2006;Haenschel, Baldeweg, Croft, Whittington, & Gruzelier, 2000). Incontrast, we found that posterior gamma activity during the stimu-lus presentation was increased for well- compared to ill-specifiedevents. Of course, the repetition of the well-specified event couldhave engendered a reliance on automatic or spontaneous retrievalmechanisms to detect intention-related events due to strengtheningof the event-to-intended action association across repetitions (Ein-stein et al., 2005). This notion is consistent with our proposal thatprestimulus activity differences between ill- and well-specifiedintentions reflect a lower engagement of attentional control proc-esses to detect events in the well-specified condition. Whether thelower reliance on demanding executive processes resulted from theexplicitly specified nature of the event or from the repeated occur-rences and detections of the specific event will be addressed insubsequent experimentation. Nevertheless, these data demonstratethat the PM event manipulation used here elicited different modesof prospective memory that were associated with differential pre-paratory neural synchronization.

Conclusion

The current findings illustrate the dynamic nature of neural mecha-nisms supporting the execution of future intentions. The differ-


ences in preparatory local and distributed neuronal synchrony,indicative of cognitive control processes (Fries, 2009; Fries et al.,2001; Sauseng et al., 2005), show that reliance upon an executivecontrol network is not identical for fulfilling intentions associatedto well- and ill-specified events. These findings provide support fortheoretical proposals that distinct attentional and mnemonic proc-

esses subserve PM depending on the type of event associated to theintention (Einstein et al., 2005; Knight et al., 2011). The flexibleexecution of future-oriented behavior in diverse environmental sce-narios may be least cognitively and neurally intensive when inten-tions are initially formulated in association with a single, specificenvironmental event.

References

Andersen, S. K., & Müller, M. M. (2010). Behavioral performance followsthe time course of neural facilitation and suppression during cued shiftsof feature-selective attention. Proceedings of the National Academy ofSciences of the United States of America, 107, 13878–13882.

Berg, P., & Scherg, M. (1994). A fast method for forward computation ofmultiple-shell spherical head models. Electroencephalography andClinical Neurophysiology, 90, 58–64.

Burgess, P. W., Gonen-Yaacovi, G., & Volle, E. (2011). Functionalneuroimaging studies of prospective memory: What have welearnt so far? Neuropsychologia, 49, 2246–2257. doi: 10.1016/j.neuropsychologia.2011.02.014

Burgess, P. W., Quayle, A. H., & Frith, C. D. (2001). Brain regions involvedin prospective memory as determined by positron emission tomography.Neuropsychologia, 39, 545–555.

Busch, N. A., & VanRullen, R. (2010). Spontaneous EEG oscillationsreveal periodic sampling of visual attention. Proceedings of theNational Academy of Sciences of the United States of America, 107,16048–16053. doi: 10.1073/pnas.1004801107

Buzsáki, G. (2006). Rhythms of the brain (p. xiv, 448). New York, NY:Oxford University Press.

Chen, Y., Huang, X., Yang, H., Ren, G., & Yue, C. (2007). Task interferencefrom event-based prospective memory:An event-related potentialsstudy.Neuroreport, 18, 1951–1955. doi: 10.1097/WNR.0b013e3282f202dd

Clementz, B. A., Gao, Y., McDowell, J. E., Moratti, S., Keedy, S. K., &Sweeney, J. A. (2010). Top-down control of visual sensory processingduring an ocular motor response inhibition task. Psychophysiology, 47,1011–1108. doi: 10.1111/j.1469-8986.2010.01026.x

Clementz, B. A., Wang, J., & Keil, A. (2008). Normal electrocorticalfacilitation but abnormal target identification during visual sustainedattention in schizophrenia. Journal of Neuroscience, 28, 13411–13418.doi: 10.1523/JNEUROSCI.4095-08.2008

Delorme, A., & Makeig, S. (2004). EEGLAB: An open source toolbox foranalysis of single-trial EEG dynamics including independent compo-nent analysis. Journal of Neuroscience Methods, 134, 9–21. doi:10.1016/j.jneumeth.2003.10.009

Einstein, G. O., McDaniel, M. A., Thomas, R., Mayfield, S., Shank, H.,Morrisette, N., & Breneiser, J. (2005). Multiple processes in prospectivememory retrieval: Factors determining monitoring versus spontaneousretrieval. Journal of Experimental Psychology: General, 134, 327–342.doi: 10.1037/0096-3445.134.3.327

Fawcett, I. P., Barnes, G. R., Hillebrand, A., & Singh, K. D. (2004). Thetemporal frequency tuning of human visual cortex investigated usingsynthetic aperture magnetometry. NeuroImage, 21, 1542–1553.

Fell, J., Ludowig, E., Staresina, B. P., Wagner, T., Kranz, T., Elger, C. E., &Axmacher, N. (2011). Medial temporal theta/alpha power enhancementprecedes successful memory encoding: Evidence based on intracranialEEG. Journal of Neuroscience, 31, 5392–5397. doi: 10.1523/JNEUROSCI.3668-10.2011

Foster, E. R., McDaniel, M. A., Repovš, G., & Hershey, T. (2009). Pro-spective memory in Parkinson disease across laboratory and self-reported everyday performance. Neuropsychology, 23, 347–358. doi:10.1037/a0014692

Fries, P. (2009). Neuronal gamma-band synchronization as a fundamentalprocess in cortical computation. Annual Review of Neuroscience, 32,209–224. doi: 10.1146/annurev.neuro.051508.135603

Fries, P., Reynolds, J. H., Rorie, A. E., & Desimone, R. (2001). Modulationof oscillatory neuronal synchronization by selective visual attention.Science, 291, 1560–1563. doi: 10.1126/science.291.5508.1560

Gilbert, S. J. (2011). Decoding the content of delayed intentions. Journalof Neuroscience, 31, 2888–2894. doi: 10.1523/JNEUROSCI.5336-10.2011

Gordon, B. A., Shelton, J. T., Bugg, J. M., McDaniel, M. A., & Head, D.(2011). Structural correlates of prospective memory. Neuropsychologia,49, 3795–3800. doi: 10.1016/j.neuropsychologia.2011.09.035

Gross, J., Kujala, J., Hamalainen, M., Timmermann, L., Schnitzler, A., &Salmelin, R. (2001). Dynamic imaging of coherent sources: Studyingneural interactions in the human brain. Proceedings of the NationalAcademy of Sciences of the United States of America, 98, 694–699. doi:10.1073/pnas.98.2.694

Gruber, T., Malinowski, P., & Müller, M. M. (2004). Modulation of oscil-latory brain activity and evoked potentials in a repetition priming task inthe human EEG. European Journal of Neuroscience, 19, 1073–1082.doi: 10.1111/j.0953-816X.2004.03176.x

Gruber, T., & Müller, M. M. (2006). Oscillatory brain activity in the humanEEG during indirect and direct memory tasks. Brain Research, 1097,194–204. doi: 10.1016/j.brainres.2006.04.069

Guderian, S., Schott, B. H., Richardson-Klavehn, A., & Düzel, E. (2009).Medial temporal theta state before an event predicts episodic encodingsuccess in humans. Proceedings of the National Academy of Sciences ofthe United States of America, 106, 5365–5370. doi: 10.1073/pnas.0900289106

Guynn, M. J. (2003). A two-process model of strategic monitoring inevent-based prospective memory: Activation/retrieval mode and check-ing. International Journal of Psychology, 38, 245–256. doi: 10.1080/00207590344000178

Haenschel, C., Baldeweg, T., Croft, R. J., Whittington, M., & Gruzelier, J.(2000). Gamma and beta frequency oscillations in response to novelauditory stimuli: A comparison of human electroencephalogram (EEG)data with in vitro models. Proceedings of the National Academy ofSciences of the United States of America, 97, 7645–7650. doi: 10.1073/pnas.120162397

Hamm, J. P., Dyckman, K. A., Ethridge, L. E., McDowell, J. E., & Clem-entz, B. A. (2010). Preparatory activations across a distributed corticalnetwork determine production of express saccades in humans. Journalof Neuroscience, 30, 7350–7357. doi: 10.1523/JNEUROSCI.0785-10.2010

Haynes, J.-D., Sakai, K., Rees, G., Gilbert, S. J., Frith, C., & Passingham,R. E. (2007). Reading hidden intentions in the human brain. CurrentBiology, 17, 323–328. doi: 10.1016/j.cub.2006.11.072

Henry, J., & MacLeod, M. (2004). A meta-analytic review of prospectivememory and aging. Psychology and Aging, 19, 27–39. doi: 10.1037/0882-7974.19.1.27

Hicks, J. L., Marsh, R. L., & Cook, G. I. (2005). Task interference intime-based, event-based, and dual intention prospective memory condi-tions. Journal of Memory and Language, 53, 430–444. doi: 10.1016/j.jml.2005.04.001

Hipp, J. F., Engel, A. K., & Siegel, M. (2011). Oscillatory synchronizationin large-scale cortical networks predicts perception. Neuron, 69, 387–396. doi: 10.1016/j.neuron.2010.12.027

Kastner, S., & Ungerleider, L. G. (2000). Mechanisms of visual attention inthe human cortex. Annual Review of Neuroscience, 23, 315–341. doi:10.1146/annurev.neuro.23.1.315

Kim, Y. J., Grabowecky, M., Paller, K. A, Muthu, K., & Suzuki, S. (2007).Attention induces synchronization-based response gain in steady-statevisual evoked potentials. Nature Neuroscience, 10, 117–125. doi:10.1038/nn1821

Klimesch, W., Sauseng, P., & Hanslmayr, S. (2007). EEG alpha oscilla-tions: The inhibition-timing hypothesis. Brain Research Reviews, 53,63–88. doi: 10.1016/j.brainresrev.2006.06.003

Knight, J. B., Ethridge, L. E., Marsh, R. L., & Clementz, B. A. (2010).Neural correlates of attentional and mnemonic processing in event-based prospective memory. Frontiers in Human Neuroscience, 4, 5. doi:10.3389/neuro.09.005.2010

Knight, J. B., Meeks, J. T., Marsh, R. L., Cook, G. I., Brewer, G. A, &Hicks, J. L. (2011). An observation on the spontaneous noticing ofprospective memory event-based cues. Journal of Experimental Psy-chology. Learning, Memory, and Cognition, 37, 298–307. doi: 10.1037/a0021969


Kucera, H., & Francis, W. N. (1967). Computational analysis of present-day American English. Providence, RI: Brown University Press.

Leber, A. B., Turk-Browne, N. B., & Chun, M. M. (2008). Neural predic-tors of moment-to-moment fluctuations in cognitive flexibility. Pro-ceedings of the National Academy of Sciences of the United States ofAmerica, 105, 13592–13597. doi: 10.1073/pnas.0805423105

Maris, E. (2012). Statistical testing in electrophysiological studies. Psycho-physiology, 49, 549–565. doi: 10.1111/j.1469-8986.2011.01320.x

Maris, E., & Oostenveld, R. (2007). Nonparametric statistical testing ofEEG- and MEG-data. Journal of Neuroscience Methods, 164, 177–190.doi: 10.1016/j.jneumeth.2007.03.024

Marsh, R. L., Hicks, J. L., Cook, G. I., Hansen, J. S., & Pallos, A. L. (2003).Interference to ongoing activities covaries with the characteristics of anevent-based intention. Journal of Experimental Psychology: Learning,Memory, and Cognition, 29, 861–870. doi: 10.1037/0278-7393.29.5.861

Meeks, J. T., & Marsh, R. L. (2010). Implementation intentions aboutnonfocal event-based prospective memory tasks. PsychologicalResearch, 74, 82–89. doi: 10.1007/s00426-008-0223-x

Miller, E. K., & Cohen, J. D. (2001). An integrative theory of prefrontalcortex function. Annual Review of Neuroscience, 24, 167–202. doi:10.1146/annurev.neuro.24.1.167

Moratti, S., Clementz, B. A., Gao, Y., Ortiz, T., & Keil, A. (2007). Neuralmechanisms of evoked oscillations: Stability and interaction with tran-sient events. Human Brain Mapping, 28, 1318–1333. doi: 10.1002/hbm.20342

Moscovitch, M. (1994). Memory and working with memory: Evaluation ofa component process model and comparisons with other models. In D.L. Schacter & E. Tulving (Eds.), Memory systems (pp. 269–310). Cam-bridge, MA: MIT Press.

Müller, M. M., Teder-Sälejärvi, W., & Hillyard, S. A. (1998). The timecourse of cortical facilitation during cued shifts of spatial attention.Nature Neuroscience, 1, 631–634.

Onton, J., & Makeig, S. (2009). High-frequency broadband modulations ofelectroencephalographic spectra. Frontiers in Human Neuroscience, 3,61. doi: 10.3389/neuro.09.061.2009

Oostenveld, R., Fries, P., Maris, E., & Schoffelen, J.-M. (2011). FieldTrip:Open source software for advanced analysis of MEG, EEG, and invasiveelectrophysiological data. Computational Intelligence and Neuro-science, 2011, 1–9. doi: 10.1155/2011/156869

Otten, L. J., Quayle, A. H., Akram, S., Ditewig, T. A., & Rugg, M. D.(2006). Brain activity before an event predicts later recollection. NatureNeuroscience, 9, 489–491. doi: 10.1038/nn1663

Praamstra, P., Kourtis, D., Kwok, H. F., & Oostenveld, R. (2006). Neuro-physiology of implicit timing in serial choice reaction-time perform-ance. The Journal of Neuroscience, 26, 5448–5455. doi: 10.1523/JNEUROSCI.0440-06.2006

Regan, D. (1989). Human brain electrophysiology: Evoked potentials andevoked magnetic fields in science and medicine. NewYork, NY: Elsevier.

Reynolds, J. R., West, R., & Braver, T. (2009). Distinct neural circuitssupport transient and sustained processes in prospective memory andworking memory. Cerebral Cortex, 19, 1208–1221. doi: 10.1093/cercor/bhn164

Sauseng, P., Klimesch, W., Doppelmayr, M., Pecherstorfer, T., Freunberger,R., & Hanslmayr, S. (2005). EEG alpha synchronization and functionalcoupling during top-down processing in a working memory task.Human Brain Mapping, 26, 148–155. doi: 10.1002/hbm.20150

Schacter, D. L., Wig, G. S., & Stevens, W. D. (2007). Reductions in corticalactivity during priming. Current Opinion in Neurobiology, 17, 171–176.doi: 10.1016/j.conb.2007.02.001

Scullin, M. K., McDaniel, M. A., & Einstein, G. O. (2010). Control of costin prospective memory: Evidence for spontaneous retrieval processes.Journal of Experimental Psychology: Learning, Memory, and Cogni-tion, 36, 190–203. doi: 10.1037/a0017732

Shulman, G. L., Pope, D. L. W., Astafiev, S. V., McAvoy, M. P., Snyder, A.Z., & Corbetta, M. (2010). Right hemisphere dominance during spatialselective attention and target detection occurs outside the dorsal fron-toparietal network. Journal of Neuroscience, 30, 3640–3651.

Simons, J. S., Schölvinck, M. L., Gilbert, S. J., Frith, C. D., & Burgess, P.W. (2006). Differential components of prospective memory? Evidencefrom fMRI. Neuropsychologia, 44, 1388–1397. doi: 10.1016/j.neuropsychologia.2006.01.005

Singer, W. (1999). Neuronal synchrony: A versatile code for the definitionof relations? Neuron, 24, 49–65. doi: 10.1016/S0896-6273(00)80821-1

Smith, R. E. (2003). The cost of remembering to remember in event-basedprospective memory: Investigating the capacity demands of delayed

intention performance. Journal of Experimental Psychology: Learning,Memory, and Cognition, 29, 347–361. doi: 10.1037/0278-7393.29.3.347

Smith, R. E., & Bayen, U. J. (2004). A multinomial model of event-basedprospective memory. Journal of Experimental Psychology: Learning,Memory, and Cognition, 30, 756–777. doi: 10.1037/0278-7393.30.4.756

Smith, R. E., Hunt, R. R., McVay, J. C., & McConnell, M. D. (2007). Thecost of event-based prospective memory: Salient target events. Journalof Experimental Psychology: Learning, Memory, and Cognition, 33,734–746.

Srinivasan, R., Bibi, F. A., & Nunez, P. L. (2006). Steady-state visualevoked potentials: Distributed local sources and wave-like dynamics aresensitive to flicker frequency. Brain Topography, 18, 167–187. doi:10.1007/s10548-006-0267-4

Srinivasan, R., Russell, D. P., Edelman, G. M., & Tononi, G. (1999).Increased synchronization of neuromagnetic responses during con-scious perception. Journal of Neuroscience, 19, 5435–5448.

Staudigl, T., Hanslmayr, S., & Bäuml, K.-H. T. (2010). Theta oscillationsreflect the dynamics of interference in episodic memory retrieval.Journal of Neuroscience, 30, 11356–11362. doi: 10.1523/JNEUROSCI.0637-10.2010

Tulving, E. (2002). Episodic memory: From mind to brain. Annual Review ofPsychology, 53, 1–25. doi: 10.1146/annurev.psych.53.100901.135114

Van Der Werf, J., Jensen, O., Fries, P., & Medendorp, W. P. (2008). Gamma-band activity in human posterior parietal cortex encodes the motor goalduring delayed prosaccades and antisaccades. Journal of Neuroscience,28, 8397–8405. doi: 10.1523/JNEUROSCI.0630-08.2008

Vialatte, F.-B., Maurice, M., Dauwels, J., & Cichocki, A. (2010). Steady-state visually evoked potentials: Focus on essential paradigms andfuture perspectives. Progress in Neurobiology, 90, 418–438. doi:10.1016/j.pneurobio.2009.11.005

Victor, J. D., & Mast, J. (1991). A new statistic for steady-state evokedpotentials. Electroencephalography and Clinical Neurophysiology, 78,378–388.

Weissman, D. H., Roberts, K. C., Visscher, K. M., & Woldorff, M. G.(2006). The neural bases of momentary lapses in attention. NatureNeuroscience, 9, 971–978. doi: 10.1038/nn1727

West, R. (2011). The temporal dynamics of prospective memory: A reviewof the ERP and prospective memory literature. Neuropsychologia, 49,2233–2245. doi: 10.1016/j.neuropsychologia.2010.12.028

West, R., Bowry, R., & Krompinger, J. (2006). The effects of workingmemory demands on the neural correlates of prospective memory. Neu-ropsychologia, 44, 197–207. doi: 10.1016/j.neuropsychologia.2005.05.003

West, R., Krompinger, J., & Bowry, R. (2005). Disruptions of preparatoryattention contribute to failures of prospective memory. PsychonomicBulletin & Review, 12, 502–507.

West, R., Scolaro, A. J., & Bailey, K. (2011). When goals collide: Theinteraction between prospective memory and task switching. CanadianJournal of Experimental Psychology, 65, 38–47. doi: 10.1037/a0022810

Woods, S. P., Twamley, E. W., Dawson, M. S., Narvaez, J. M., & Jeste, D.V. (2007). Deficits in cue detection and intention retrieval underlieprospective memory impairment in schizophrenia. SchizophreniaResearch, 90, 344–350. doi: 10.1016/j.schres.2006.11.005

Worsley, K. J. (2003). Detecting activation in fMRI data. StatisticalMethods in Medical Research, 12, 401–418. doi: 10.1191/0962280203sm340ra

(Received May 10, 2012; Accepted May 12, 2012)

Supporting Information

Additional supporting information may be found in the onlineversion of this article:

Appendix S1: Supplemental results

Please note: Wiley-Blackwell are not responsible for the content orfunctionality of any supporting materials supplied by the authors.Any queries (other than missing material) should be directed to thecorresponding author for the article.


preparatory distributed cortical synchronization ...gbrewer1/http___~gbrewer1... · preparatory...

Documents